AI prompts for marketers

Before any campaign: what does this audience actually complain about, in their own words?

Below is raw audience material about [CATEGORY/PRODUCT]: reviews, support tickets, forum threads, social comments.

Extract and deliver:
1) Top 7 pains, each with 2 verbatim quotes (exact wording, no cleanup).
2) Top 5 desired outcomes — their words, not marketing terms.
3) The emotional undertone per pain (frustration / anxiety / resignation…).
4) 5 message hypotheses, each pairing one pain quote with one outcome quote.

Do not invent anything not present in the material. If a category is thin, say so.

Material:
[PASTE 30–200 SNIPPETS]

You need genuinely different campaign directions to react to — not five flavors of the same safe idea.

Product: [PRODUCT]. Audience: [SPECIFIC SEGMENT]. Core fact: [THE ONE QUANTIFIED BENEFIT]. Budget class: [E.G. SMALL/SOCIAL-ONLY]. Brand register: [3 ADJECTIVES + TABOOS].

Generate 6 campaign concepts, one per lever:
1) Magnify — dramatize the core fact to its extreme.
2) Reverse — flip an industry cliché on its head.
3) Substitute — tell it as a customer's story, no product claims.
4) Combine — merge the benefit with the biggest objection.
5) Minify — the 3-word version.
6) Adapt — borrow a mechanic from another industry: [OPTIONAL: NAME ONE].

Per concept: name, one-line big idea, hero execution, first headline. The core fact must appear in at least 4 concepts.

Concept chosen. Now the boring-but-critical part: what runs where, when, with what message depth.

You are my [performance marketing lead]. Build a channel plan skeleton for the campaign below.

Campaign: [CONCEPT ONE-LINER]. Objective + KPI: [E.G. 500 SIGNUPS, CPL < €X]. Budget: [TOTAL + SPLIT CONSTRAINTS]. Duration: [WEEKS]. Audience lives on: [WHAT YOU KNOW].

Format: table — channel, role in funnel, message angle (from the concept), format specs, budget %, primary metric, kill criterion (when we stop spending). Then: 3 assumptions this plan makes that we should validate in week 1, each with the cheapest validation test.

Ask me up to 3 questions before building if anything is missing.

Month's over, dashboard is full, stakeholders want to know what happened — in one page, in sentences.

Turn the metrics below into a performance readout for [AUDIENCE, E.G. CMO / CLIENT].

Format: 1) Three-sentence summary: what happened, why, what we're changing. 2) Wins and losses — max 3 each, every claim with its number and comparison base (vs. last period / vs. target). 3) One "honest paragraph": what we don't understand yet. 4) Next month's 3 changes, each tied to a finding above.

Rules: no metric without a comparison, no "strong performance" without the number, mark anything the data doesn't support as [HYPOTHESIS].

Data:
[PASTE METRICS TABLE / EXPORT]
Context: [WHAT CHANGED THIS PERIOD — LAUNCHES, BUDGET SHIFTS, SEASONALITY]

One landing page or ad set, and you want a prioritized month of tests instead of random tweaks.

Here is our current [LANDING PAGE COPY / AD SET] and its numbers: [PASTE ASSET + CURRENT CTR/CVR/CPL].

Generate a test backlog: 8 A/B test ideas, each with — hypothesis ("changing X to Y will improve Z because…"), the mechanic behind it (clarity / proof / urgency / friction / specificity), effort (S/M/L), and expected signal strength given our traffic of [SESSIONS OR IMPRESSIONS / WEEK].

Prioritize into: run now (2), run next (3), park (3) — justify the top 2 in one line each. Base every hypothesis on something visible in the asset or numbers, not general best practice.